2 papers accepted for: Workshop on Information in Networks
September 28 - 29, 2012
Stern School, NYU
Genc, Y., Mason, W., and Nickerson, J.V.
Semantic Transforms Using Collaborative Knowledge Bases
As the amount of textual information increases in the web, it becomes increasingly important to be able to automatically understand the content of the text; classifying it into specific categories accurately. Primarily this is accomplished with topic models, particularly latent Dirichlet allocation (LDA). However, while topic models find sets of keywords, they do not provide a concise human-understandable topic designation for a document: usually, humans create one by looking at the keywords produced by a model. In this work, we create variants of topic models in which documents are classified by the title of the Wikipedia page that they match best. This model has many potential advantages over traditional topic models, because it can take advantage of human-understandable topic names – Wikipedia page titles – to describe texts.
Kyriako, H, Englehardt, S. and Nickerson, J. V.
Networks of Innovation in 3D Printing
Innovation inside companies is difficult to see. But an emerging online community of inventors who publicly post 3D CAD drawings of their work provide a way to observe – and perhaps amplify – innovation. In this paper we analyze the network structure of Thingiverse, a website oriented toward 3D printing. This form of printing blurs the line between creating information and manufacturing objects: drawings can be sent to devices that build 3D objects out of many materials, including resin, ceramics, and metal. In this exploratory study we analyzed the structure of Thingiverse links. Our results suggest that analysis of remix network structure may provide ways of tracing innovation processes and detecting the emergence of new ideas, combination of disparate ideas.